Max Robotics — Global Service Robot Directory + US Certification

About

David Held is a robotics and computer vision researcher whose work spans 3D perception, robot learning, and safe reinforcement learning. He is perhaps best known for developing the Point Completion Network (PCN), a landmark contribution to 3D shape completion that estimates full object geometry from partial observations — a paper that has accumulated nearly 1,000 citations and become a foundational reference in the field. His influential work on 3D Multi-Object Tracking, which introduced practical baselines and new evaluation metrics for autonomous driving and assistive robotics applications, has garnered nearly 500 citations, reflecting its broad adoption by the research community. Held has made significant strides in reinforcement learning as well, contributing the widely-cited Reverse Curriculum Generation method for goal-oriented robot tasks and co-developing Constrained Policy Optimization, an approach enabling safer RL systems that respect explicit behavioral constraints — particularly relevant for human-robot interaction. His research further extends to deformable object manipulation through the SoftGym benchmark and to communicating robot intent to human collaborators. Across these diverse contributions, Held's work consistently bridges perception, learning, and safety, making him a prominent voice in modern robotics research.